Re: Pentium DEATH in user-mode

Benson L Chow (bc3c+@andrew.cmu.edu)
Sat, 8 Nov 1997 10:53:37 -0500 (EST)

Uh... I'd imagine this would take a long time to search every single op
code. This is because of the recovery needed if you hit like a branch
or something that modifies control flow... I guess you could use single
step mode but you'd take an interrupt on each instruction, this would
take quite a long time. Must be a really bored person to go through
this.

Even if it didn't crash in 32 bits, who knows what would happen in two
instructions, etc., such as the 386's popad bug...

Morale: Time to replace p5's in multiuser servers with something else...

Excerpts from internet.computing.linux-kernel: 8-Nov-97 Re: Pentium
DEATH in user-mode by Andre Derrick Balsa@usa.
> > Run the program to search all of the ops (modify it to have it search all
> > 32-bits of 'em) and see if that'll crash the cyrix.
> >
> Done it already. No problems. :-)
>
> Cyrix: 2^32, Intel: -0.000712765430978 ;-)
> Cheers